Salient object detection aims to locate objects that capture viewer's attention within images. Previous approaches often exploit multiple priors to improve the saliency detection results. In this letter we present a novel saliency method by applying two simple priors namely color-spatial, and object proposal. We note directly applying existing these priors does not produce reasonable results. Our solution is to establish the connection between these modified priors via an optimization technique, where a novel, simply iterative manner is introduced as a reliable support for saliency detection. Our framework can produce more accurate saliency maps, yet performs favorably against several state-of-the-art methods on the three image datasets.
展开▼
